Volume 17 Number 13 1989

Nucleic Acids Research

The isochore organization and the compositional distribution of homologous coding sequences in
the nuclear genome of plants

Giorgio Matassi, Luis M.Monterol, Julio Salinas' and Giorgio Bernardi*
Laboratoire de Gdndtique Moldculaire, Institut Jacques Monod, 2 Place Jussieu, 75005 Paris, France
and 'Departamento de Proteccion Vegetal, Instituto Nacional de Investigaciones Agrarias, Carretera de
la Coruna, Km. 7, 28040 Madrid, Spain
Received March 28, 1989; Revised and Accepted June 2, 1989

ABSTRACT

The isochore structure of the nuclear genome of angiosperms
described by Salinas et al. (1) was confirmed by using a
different experimental approach, namely by showing that the
levels of coding sequences from both dicots and Gramineae are
linearly correlated with GC levels of the corresponding flanking
sequences. The compositional distribution of homologous coding
sequences from several orders of dicots and from Gramineae were
also studied and shown to mimick the compositional distributions
previously seen (l)for coding sequences in general, most coding
sequences from Gramineae being much higher than those of the
dicots explored. Thfese differences were even stronger for third
codon positions and led to striking codon usages for many coding
sequences especially in the case of Gramineae.
INTRODUCTION
Recent investigations on plant genomes (1) have led to two
major conclusions : (i) that the nuclear genomes of angiosperms
are mosaics of long (>100-200 Kb), compositionally homogeneous
DNA segments, which were called isochores (2), and that belong
to families characterized by different GC levels; and (ii) that
the nuclear genomes of Gramineae exhibit strikingly different
isochore patterns compared to those of the dicots investigated.
Differences in isochore patterns were seen at two different
levels. (i) When DNA fragments (in the 50-100 Kb range)
fromthree dicots (pea, sunflower and tobacco) were compared with
those from three Gramineae (maize, rice and wheat), the
compositional distribution of the fragments was centered around
41% GC for the former and around 45% GC for the latter.
Moreover, in the case of maize and wheat, distributions trailed
towards
even higher GC values. (ii) Coding sequences from

© I R L Press

5273

Nucleic Acids Research
several orders of dicots showed a narrow, symmetrical
compositional distribution centered around 46% GC. In contrast,
coding sequences from barley, maize and wheat showed a broad,
asymmetrical distribution characterized by an upward trend
towards high GC values, the majority of sequences being
comprised between 60 and 70% GC. Similar, yet more striking,
differences were found when GC levels of third codon positions
introns exhibited compositional
were compared. Finally,
distributions that mimicked those of exons from the same genes,
but were lower in GC levels, as were the intergenic non-coding
sequences that form the vast majority of plant DNAs. The latter
point is indicated by the lower GC levels of DNA fragments
compared to coding sequences from the same genomes.
To sum up, the genome organization of angiosperms is very
reminiscent of that previously described for vertebrates (2),
with a striking and puzzling resemblance in compositional
patterns between the dicots investigated and cold-blooded
vertebrates, on the one hand, and of Gramineae and warm-blooded
vertebrates, on the other. In the present work, we have
investigated the compositional distributions of DNA fragments in
the 50-100 Kb size range for two additional dicots and six
additional monocots. We have then compared the GC levels of
homologous coding sequences. (and of their different codon
positions) from several dicots and from Gramineae, in order to
better define the differences in compositional patterns between
the genomes of these plants, and the consequences of such
differences on codon usage. Finally, we have investigated the
relationships of GC levels of coding sequences and introns with
those of the corresponding flanking sequences.

METHODS

Preparation and fractionation of nuclear

DNAs.

Ethiolated

seedlings from Hordeum vulgare(barley), Secale cereale(rye), and
Asprgs officinalis, and mature leaves from Antirrhinum majus,
Oenothera hookeri, Scindapsus aureus, Ty2ha latifolia, and
Allium ce2a(onion), were used to prepare nuclear DNA as
described elsewhere (1). Details on the characterization of the

5274

Nucleic Acids Research
DNAs and their fractions will be presented in a forthcoming
paper.
GC levels of coding sequences (from the initial AUG to the
terminal codon), of first, second and third codon positions, and
of introns and flanking sequences were obtained from GenBank
(Release 57, September 1988) . Data concerning genes not yet
available in GenBank were obtained from the literature. A
reference list for the coding sequences from the literature and
a list of the genes studied in both exons and introns will be
provided upon request. The ACNUC retrieval system (3) was used.
Codon_usaje was investigated by considering the ratio of
observed codon frequency over codon frequency expected for a
statistical distribution of codons (this ratio has been called
relative synonymous codon usage, or RSCU; see ref. 4).
The homologous genes investigated, the plants from which
they derived, and the plant families were the following :
Actin
Glycine max (Le9uminosae, 2 genes); Zea
mnays(Gramineae) .Alcohol dehydrogenase (Adh)
Arabido2sis
thaliana (Cruciferae); Pisum sativum (Leiuminosae).
Chlorophyll-a/b-binding protein(Cab)
A. thaliana (3 genes);
Cucurbita sp. (Cucurbitaceae; 2 genes); Lyogersicon esculentum
(Solanaceae; 2 genes); Petunia_hbrida (Solanaceae; 6 genes); P.
sativum; Silene _ratensis (Caryophillaceae) ; Lemna _ibba
(Lemnaceae); Triticum aestivum (Gramineae); Z. may.Chalcone
synthase(Chs): Antirrhinum majus (Scrofulariaceae); Maanolia

liliiflora (Mjnoliaceae); Petroselium hortense (Umbelliferae);
h_Zrida; Phaseolus vuga ris (Leguminosae); Ranunculus acer
(Ranuncolaceae); Hordeum vulaare (Gramineae); Z. mays
Glyceraldehyde-3-phosphate-dehydrogenase (GADPH)
Sinapis alba
P.

(Cruciferae, 2 genes); Nicotiana tabacum (Solanaceae, 3 genes);
Z. mays.Histone H3(H3): A. thaliana; rza sativa (Gramineae) ;
T. aestivum; Z. mays (2 genes) . Histone H4 (H4) : A. thaliana;
T. aestivum; Z. mays (2 genes) .Phytochrome(Phyt.)
Cucurbita
*
Avena
sativa
(Gramineae, 2 genes) .
2e2o

Ribulose-l,5-bisphosphate carboxylase, small subunit (Rubisco)
Cucurbita sp.; G. max (2 genes); Helianthus annuus (Compositae);
L. ibba; L. esculentum (3 genes); N. tabacum; N. sylvestris

5275

Nucleic Acids Research
(Solanaceae); O. sativa; P. _hbrida (3 genes); P. sativum (4
genes); Raphanus sativus (Cruciferae); S. Eratensis; T.

aestivum; Z. mays.
RESULTS

Compositional distribution of DNA fragnts
Fig. lshows histograms providing the buoyant densities in
CsCl and the relative amounts of compositional fractions as
obtained from CsIS04/BAMD fractionation of nuclear DNAs from
five dicots and nine monocots (BAMD is 3,6 bis
(acetatomercurimethyl) dioxane).
As far as dicots are concerned, two additional DNAs (from
Antirrhinum majus and Oenothera hookeri, respectively) were

D

a

cC

Buoyant density

,

g/cm3

Fit. 1. Histograms showing the relative amounts and buoyant
densities in CsCl of DNA fractions obtained by preparative Cs SO
/BAMD density gradient centrifugation from dicots (left panel) and
monocots (middle and right panels) . Data for pea, sunflower,
tobacco, maize, rice and wheat are from ref.l (where additional
details can be found). The vertical broken line at 1.700 g/cm3 is
shown to provide a reference. Black bars are used whenever
required for discriminating fractions from different DNAs.
Horizontal lines on some bars separate DNA fractions showing the
same buoyant densities.
5276

Nucleic Acids Research
studied compared to our previous investigations (1) . These DNAs
were chosen because literature data (5,6) indicated that they
corresponded to the lowest and highest buoyant densities of all
dicot DNAs investigated so far.
The new monocot DNAs studied were from two additional
Gramineae, rye and barley, and from four other monocots,
Scinda2sus aureus, TyEha latifolia, Allium ce2a (onion) and
As2araaus officinalis. Rye and barley were chosen because a
number of their coding sequences are known in primary structure.
The choice of the other four monocots was made in order to cover
a larger spectrum of families (Scindapsus belongs to Araceae,
Coding sequences

4
2

Actin

o

m

2
0

m, *

Adh

Cab

L

c2

Chs

2

GAPDH

O 0G~~~~~Histones
f

@

0

4

E2
6 ~~
A

~Ru bisco

0

40

50

60
GC ,.%

70

80

F _2.The numbers of homologous genes are plotted against
the GC levels of the corresponding coding sequences; a 2% GC
window was used. Values for dicots and monocots (Gramineae,
except for two genes, cab and rubisco, from L. _ibba) are
represented by the open and black bar histograms, respectively.
Genes for histones H3 and H4 are presented in the same
histogram. H3 genes from Gramineae are indicated by horizontally
hatched bars, their homologs from dicots (Arabidopsis) by
vertically hatched bars. See Materials and Methods.
5277

Nucleic Acids Research
4
2

First codon poi tion
Act n
O

0
2

Ad h

0
Cob
4)
c
4)
0)
L

-o

E
z

4
2
0
2
0
2
0
2
0
2
0
8
6
4
2
0

r-

Chs

n

nF l-

-GAPDH
His tones

R u b sc o

40

A
:~~~~~~~~~~~~~~~~~~~~~~60

50

I

70e,%

GC, %

Fij. 3.The numbers of homologous genes of Fig. 2 are
plotted against the GC levels of the first codon position of the
corresponding coding sequences. The highest value among cab
genes belongs to Silene pratensis. Other indications as in Fig.
2.
to Typhaceae, Asparagus to AspaEraaceae and Allium to
Alliaceae) compared to those previously explored, and to analyze
a DNA (that of Allium) characterized by one of the lowest modal
buoyant densities among plant DNAs (5).
The data of Fig. 1 stress the fact that DNAs from both
dicots and monocots cover a wide range of modal buoyant
densities (namely, of the buoyant densities at the peak of
analytical CsCl profiles). In the case of the dicots studied
here, the limits of the range are 1.6942 (Antirrhinum_mjus) and
1.7035 g/cm3 (Oenothera hookeri). In the case of monocots, the
range of modal buoyant densities is wider, from 1.691 g/cm3 for
Allium_ce2a to 1.7026 g/cm3 for T. aestivum; all values for
Gramineae are in a higher narrow range, 1.7107 to 1.7026 g/cm3.

Tyha

5278

Nucleic Acids Research
Second codon position
4
2
0
2
0

4
2

t0
U,,O
2

0 o

E04
2
0

8
6
4
2

0
GC ,%

Fij. 4. The numbers of homologous genes of Fig. 2 are

plotted against the GC levels of the second codon position of

the corresponding coding sequences. The highest value among cab
genes belongs to Cucumis sativus. Other indications as in Fig.
2.

Compositional distribution

codinm sequences from
dicots and Gramineae
Fig. 2 displays the compositional distribution for all
available homologous coding sequences from dicots and monocots.
As indicated in the Materials and Methods section, the nine sets
of homologous genes investigated were from several dicots,
belonging to ten different families, but from only five
monocots, four of which belonged to a single family , Gramineae
(only two genes, cab and rubisco, were from another monocot,
Lemna_jibba). GC levels of coding sequences were higher in
monocots than in dicots for all nine genes tested. When
comparisons were made for first, second and third codon
positions (Figs. 3-5), the differences found in coding sequences
of homologous

5279

Nucleic Acids Research
Third codon

A c t in

2

Ad h

C~~~
,0 50
2

E 0
L

position

4
2

Cab

_

mnn*
GA PD H

z2

2

Phy tochrome

6
o30 n, l,50
b
~~~~
|
g
~~~~Ru
2 2

30

40

50

C60/7

80

sco

90 100

Fij.5. The numbers of homologous genes of Fig. 2 are
plotted against the GC levels of the third codon position of the
corresponding coding sequences. Other indications as in Fig. 2.
became much larger when third codon positions were compared,
whereas they became smaller for first codon positions, and
almost disappeared for second codon positions.
Codon

usage

The codon usage of all genes shown in Fig. 2 was analysed.
Table 1 displays an example of tis analysis for nine pairs of
homologous genes from dicots and monocots, respectively. In the
case of monocots, Zeamw
was presented systematically (except
for the phytochrome gene which is not available), in order to
provide information on codon usage patterns within a single
species. These results will be commented upon in the Discussion.

CoTearison

of

GC

levels

of

exons

and

introns

with

GC

levels

of

The GC levels of coding sequences were compared with those
of the corresponding flanking sequences. In this case, the gene
sample used was largely different from that of homologous genes
and comprised all coding sequences (from the GenBank and the
literature) for which pooled 5'and 3' flanking sequences were
longer than 1 Kb. A straight-line relationship with a slope of
5280

Nucleic Acids Research

Phe TST

TIC

Zs
As
Cp
Phyt. Phyt. Actl

GM
Actl

Zm

Ps

Adhl

AdhI

CAPDIO

SO.
CAPDH

1.14
0.86

0.91
1.09

0.50
1.50

0.72
1.28

0.43
1.57

0.57
1.43

0.20
1.24
1.86

0.30

0.50

0.30
1.80
0.90

1.00
2.75
1.00

0.30
1.S0

1.43
1.72

3.00

1.66

2.70

0.75

1.20

2.04

0.99
1.50

0.82

0.72
U.24

0.73
2.20
0.13

1.24

0.73

0.76

1.27
1.15
1.39
1.85
0.46
1.15

CTG

1.21

1.38
2.05
1.14
0.29
0.62
0.52

Ile ATT
ATC
ATA

1.07
1.34
0.59

1.48
0.87
0.65

1.50
1.37
0.13

Val GTT
GTC
CTA
GTG

1.40

0.77
0.72
1.11

1.43
0.64
0.84
1.09

0.39
0.64

2.06
0.52
0.39
1.03

0.92
0.92
0. 31
1.85

Ser TCT
TCC
TCA

1.28
1.21
0.94
0.40

1.38
0.48

0.64
1.28

1.28
0.86

0.35
0.35

Leu TTA
TTG
CTT

0.34
1.11
1.60

CTC
CTA

0.82
0.92

TCG

1.56
0.52
1.65

1.S5
0.41

1.42
1.55

0.43

0.83
0.20

2.18

Zs
13

1.29

0.86

0.38

1.71 3.00
m

2.14

2.62

0.50

0.57
1.99

1.99
1.14

0 .49

1. 72
2.14

0.86
0.43

4.15

0.86

J.08
0.16
2.75

1.00
2.00

3.00

1.46
0. 36
0.36
1.82

1.89

1. 20
0 .40

2.81

1.47

1.00
2.50
1.00

3.14

1.11
1.11
.67
1.11

0.29

_

1.33
1.33
U.:3
0 3

2.50
0.50

2.00
1.33

2.00

1.00
2.00

0.24
2.44

1.00
0 .40

2. 18

0.67

2.00

1.00

1.56

0.80

1.82

0.50
1.75
0.25
1.00

1.36
1.09

1.06
1.76

1.14
1.43
1.14
0.29

2.33

2.21
0.95
0.63
0.32

2.25

1.90
0.21
1.47
0.42

1.20
0.60
0.80
1.40

1.20
0.40
1.60
0.80

1.33
0.67
0.67

1.33
0.44
2.22

1.67

1.82

1.80
1.80
0.20
0.20

1.40
1.00
0.8U

1.09
2.00
0.73
0.18

1.64
1.46
0.54
O.J6

1.41
0.94
0.82
0.82

2.34
0.28
1.10
0.28

1.93
1.79
0.28

2.52
1.04
0.44

1.58

1.33

0.82

3.00

2.40

2.40

1.09
1.09

0.67
2.67

1.33

0.67

1.33
1.33

4.00

2.40

1.20
2.00
0.80

4.00

1.24
0.76

1.52
0.48

0.67

1.29
0.71

0.40
1.60

0.33
1.67

0.67

0.50

1.33

1.S0

2.00

1.00
1.00

0.80

_

1.62
0.50

1.18

0.12

1.00
1.00

_

_

2.00

2.00

2.00

1.43

2.00

1.00
1.00

2.00

2.00

2.00

0.67
1.33

2.00

0.75
1.25

1.56

0.75

0.44

0.91
1.09

1.33

1.20

0.67

1.25

1.09

0.40
1.60

0.57
1.43

2.00

Asn AAT
AAC

0.93

1.50
0.50

0.22
1.78

1.43

0.44
1.56

1.07
0.93

0.46

1.07

1.54

1.60

Lys AAA

8.50
1.50

0.86
1.14

0.21
1.79

0.42
1.58

0.38
1.62

0.92
1.08

0.29
1.71

0.32
1.68

2.00

1.29
0.71

1.37
0.63

1.39

1.36
0.64

0.77
1.23

1.68
0.40

0.S4

0.68

0.61

1.46

1.12

0.93
1.07

0.27
1.73

0.24
1.76

2.00
0.55

1.43

0.64
1.36

Cys TGT

1.13
0.87

1.40
0.60

0.80
1.20

1.33
0.67

0.62
1.38

1.64
0.46

Arg CGT
CGC
CGA
CGG

0.68

1.11
0.22
0.89
0.34

1.33
0.34
0.34
0.34

4.24
2.57

0.92
0.46
0.92

0.35

0.86

Set ACT
AGC

1.10

1.38
0.76

0.64
2.36

0.43
1.93

0.35

0.29

2.12

1.72

_

2.00
2.00

2.00
2.00

1.75

0.21
1.90

1.26
0.63
1.47
0.63

2.77

0.92
1.23

0.21

1.68

0.77

0.33
3.33

1.23

2.04

1.96

0.57
1.6U
0.92
0.92

2.00

2.00

2.00

0.67
1.33

2.00

2.00

2.00

0.40
1.60

2.00

1.14
0.86

2.00

0.91
1.09

0.33

2.00
2.00

1.08

2.15
0.92
0.92

0.57

2.00

0.40

0.91
2.00

2.00

2.00

1.09

2.00

0.29
1.71

2.00

1.09
0.91

1.29

2.00

0.20
1.80

2.00

0.29
1.71

0.20
1.80

2.00

0.10
1.90

0.38
1.62

0.50o
1.50o

1.S0
0.50

2.00

1.33
0.67

2.00

0.91
1.09

2.00

1.14
0.86

2.10

0.83
1.17

0.13

2.00

1.14
0.86

2.00

2.00

0.57
1.43

2.00

0.67
1.33

2.00

0.45
1.55

2.00

2.00

2.00

U

_
_m
2.00

2.00

1.00

1.50

0.35
4.59

1.76
0.35

4.00

1.99
1.33

1.09

2.OU
0.71

_ _

1.09
1.09

_

1.87

_

_

2.00

2.00

_

0.40

2.00

2.57

2.40
8
0.80

0.35
2.50

0.24
2.70

2.67
1.33

_

0.44
1.56

0.57

0.80
0.80

2.22

2.28
1.14

1.09
2.18

0.30
2.40
0 .30

0 .93

1.72

0.73

2.40

1.78

0.57

0.80

1.14
0.86

2.28
1.71

1.50

2.33

_

1.29
0.71

0.99
1.01

1.05
0.42
2.53

2.28

1.52
0.48

Glu GAA
GAG

4.00

1.20

1.60

0.20
1.80

0.73
1.27

0.47
1.53

1.00

0.57
1.14

1.29
0.71

0.97

3.00

1.80

1.20
0.60
1.00

0.40

His CAT
CAC

2.00

3.00

1.60

Gln CAA
CAG

0.91

1.28

1.20

0.55

0.49
1 .30
1.78
0.65
1.30

1.85

1.33

Tyr TAT
TAC

3rd

0.77
1.23

2.67

1.16
0.42

G+CS

2.00

0.65

0.58

1.39

2.00

0.44

0.29

2.15

M1
Chs

2.16
1.89

0.75
1.21

AGG

Zm
Chs

0.45
1.55

1.89

1.52
0.96
1.52

Gly GGT

Rub.

1.67

1.67
1.67
0.17
8.50

1.58

Gm

Rub.

1.89
0.76

1.84

Arg AGA

Cm

1.94

1.75

1.10

2.08

0.43
2.57

Ala GCT
GCC
GCA
GCG

8.56
0.91

0.75

1.50

0.38

0.11

3.23

1.00

0.91

TGC

3.00

2.00

0.95
1.33
1.33

CAC

0.23
0.46

0.68
2.32

0.73

Asp GAT

1.50
0.75

4.50

1.72
0.48
1.52
0.28

AAG

2.00

1.06
0.94

4.00

1.30
1.02
1.40
0.28

1.33

Ph
Cab

1.50
1.99
1.00
0.50

Tht ACT
ACC
ACA
ACC

1.09
1.27

Cab

2.87
0.29
0.29

0.26

1.14
0.38

Zs

114

1.00
1.00

1.60

0.57

CCC
CCA
CCG

At

114

1.00
1.00

2.0U

2.24
0.49
0.98
0.29

Pro CCT

Z2

At
HJ

0.40

0 11

0.64
1.90

Zs

0.26
3.65

2.63

_

130

1.13
0.75

_
2.25

_
1.50

1.14

1.55

1.72
0. 86

0.67
0. 67

0.67

_
0. 78

0.75
0. 75

2.86
0.34
0. 80

0 .78

0.78

5.25
_

4.00

0.75

0786

2.66

0.63
1.27

_

3.00

0.67

1.20

3.60

_

1.76
2.12

1.20

1.20
1.99

0.57
3.43

0.47
2.59
0.24
0.78

1.65
0.24
2.12

0.12
3.12
0 12
0.62

1.88
0.38
1. 50
0. 25

3.27
0.36
0. 36

_
_

1.00

0.44

1.99
1.45

1.99
1.67

0.71

0.43

0.71

2.14

0.92

3.27

1.60
1.07
0.53
0.80

1.62
0.50

1.62

1.84

1.87
1.47
0.40
0.27

0.28
0.26

0.38

0.97
0.54
8.86

0.32
0.86
0.97

45.9

54.0

48.6

68.6

51.1

61.5

51.9

69.4

50.1

64.6

50.8

69.3

53.7

41.7

68.7

56.0

93.4

50.4

94.2

53.9

95.6

45.7

97.0

61.4

97.7

62.1

GGC

0.64

GGA
GGG

1.12

1.01
0.64
1.39

0.75

0.96

1.Su

47.1

41.9

51.4

47.0

54.9

46.8

37.7

59.6

42.9

66.6

2.77

1.09

2.50
1.99

0.71

1.86

0.57
2.86
0.57

1.99
1.14
1.43
1.43

1.29
1.16

Table 1. Codon usage exhibited by ninepairs of homologous genes
from monocots and dicots, res2ectivey. Values are ratios of
observed codon frequency over codon frequency expected for a
statistical distribution of codons (RSCU; ref. 4). Met, Trp and
termination codons were not taken into account. Dashes stand for
absent codons. Gene pairs were ordered from left to right
according to increasing GC content in third codon position of
genes from monocots (A. sativa and Z. mats). The two bottom
lines present GC levels of coding sequences and third codon

positions.
5281

Nucleic Acids Research

20
5

I ntrons

80
60_

20

50
40
30
Flanking sequences,GCY%

60

Fi._6. GC levels of exons (upper frame) or introns (lower
frame) are plotted against the GC levels of the corresponding
flanking sequences. Open and closed points correspond to dicots
and monocots (Gramineae except for one point from L. _ibba,
respectively. The sequence list will be provided upon request.
0.80 and a correlation coefficient of 0.72 was obtained by using
the least square procedure (Fig. 6) . Points for genes from
dicots corresponded to lower GC values and were less scattered
compared to those of Gramineae, (except for one point
corresponding to Lemna). When all available 5' and 3' flanking
sequences (regardless of size) were plotted separately, results
were similar. Slopes were 0.77 (5') and 0.73 (3') with
correlation coefficients of 0.73 and 0.60, respectively (not
shown). Fig.6 also indicates that the flanking sequences of
Gramineae are characterized by higher GC levels than those of
dicots.
When GC levels of introns from the same genes were plotted
against GC levels of flanking sequences, points were less
scattered and fell on a straight line with a slope of 0.86 and a
correlation coefficient of 0.82. The straight line of introns
was displaced towards the bottom of the diagram compared to that
of coding sequences (Fig. 6).
If GC levels of individual codon positions of the same
genes were plotted against GC levels of corresponding flanking
sequences, points showed a larger scatter compared to the plot
5282

Nucleic Acids Research
100

position

Ist codon
80

0~~*0~

6

60

O00

40
20

position

2nd codon

80
60

0
0

U

00

v40
20

.

3rd codon

80

.

posi t ion

60

_
0

0

40

4
20

0

0

A

30
40
60
50
Flanking sequences,GC/.

Fig. 7. GC levels of codon positions from the genes of Fig.
6 are plotted against the GC levels of the corresponding
flanking sequences. The highest GC value in third codon position
is the cab gene from L. qibba. See legend of Fig. 6 for other
indications.
of coding sequences (Fig. 7). Slopes ranged from 0.1 for the
second, to 0.52 for the first and to 1.78 for the third codon
position respectively (Fig. 7) and correlation coefficients were
0.11, 0.47 and 0.70, respectively.
DISCUSSION

Co2Eositional

distributions of DNA

fraqments

from dicots and

monocots

The range of modal buoyant densities of DNAs from dicots
deserves several comments. (i) The range of all dicots
investigated so far is narrower, 1.694-1.697 g/cm3, than that of
monocots, with the only exception of Oenothera. Needless to say,
exploration of DNAs from other dicots may reveal that the range
is broader than that seen at present. (ii) The lowest value of
dicot DNAs found here, that of Antirrhinum, is remarkably higher
5283

Nucleic Acids Research
than previously reported, 1.691 g/cm3 (5). Antirrhinum DNA
exhibits, however, a remarkable heterogeneity with a very
GC-poor component (possibly a satellite) as low as 1.6895 g/cm3
in buoyant density. Unfortunately, no data are available on the
methylation of this DNA, a fact preventing any precise
assessment of its GC content. Since methylation leads to a
decrease in buoyant density of about 0.7 mg/cm3 per 1t 5-methyl
cytosine (1), the GC content of Antirrhinum DNA (which would be
about 34% in the absence of methylation) can only be
underestimated if calculated from its buoyant density. (iii) The
upper limit of modal buoyant densities of dicot DNAs is that of
Oenothera. Our value for modal buoyant density, 1.7035 g/cm3, is
in good agreement with previous ones for DNAs from several
Oenothera species, 1.703 g/cm3 (5). Our Cs.SO4analysis (Fig. 1)
has shown, however, the existence of a remarkably lighter peak,
1.6972 g/cm3, representing as much as 30% of total DNA. While
the latter is likely to correspond to a satellite, this needs to
be established. Again, since no information is available on the
methylation of Oenothera DNA, no estimate of its GC content can
be given. In conclusion, the GC levels of the DNAs from dicots
investigated so far range from a minimal value of 34% to a value
of 43% (if the 1.703 g/cm3 peak is indeed the main band) or
higher if Oenothera DNA is methylated.
In the case of monocot DNAs, two points should be made. (i)
Since methylation data are available for both onion and wheat
DNA (6) , the GC range of the DNAs studied can be estimated as
comprised between 38 and 48% (1). (ii) An interesting
observation concerns the difference in the compositional
patterns exhibited by Aliumcepa and Aspagus officinalis, two
species belonging to the same order,Liliales. The corresponding
DNAs differ by as much as 5.3 mg/cm3 in modal buoyant density,
and buoyant density distributions show no overlap. This
situation stresses the possible lack of correlation between
taxonomy and genome compositional patterns and is similar to
that found for a number of fish orders and families (G.Bernardi
and G.Bernardi, paper in preparation; see also ref. 8).
The expanded set of plant genomes analyzed indicates, in
agreement with previous literature data (5,6), that both dicots
5284

Nucleic Acids Research
and monocots cover a wide range of GC contents, which does not
seem to differ very much in its extreme values.
Cor2arisons of homoloqous coding_ seuences from dicots and
Gramineae
Fig. 2 shows that the GC levels of the coding sequences
from the dicots analyzed range from 40 to 56% GC, with most
values comprised between 44 and 54%. In other words, they
correspond to the center of the compositional distribution
previously seen for coding sequences from dicots (1). The coding
sequences from Gramineae cover a much broader range, from 44 to
70% GC, most values being, however, in the 60-70% GC range. This
distribution also largely reflects the general distribution of
coding sequences from Gramineae, which comprises both a low GC
and a high GC range, with most coding sequences in the latter

(1).
When third codon positions of homologous genes are
examined, the GC range covered in the case of dicots was 36-64%,
which again, expectedly, corresponds to the center of
distribution of third codon positions of all available sequences
for dicots (1). In the case of Gramineae, the range was 46-100%;
this covered both the low and the high GC range of all third
codon positions. Expectedly again, the lowest and the highest GC
values corresponded to the coding sequences which were lowest
and highest in GC, respectively.
Second codon positions showed GC levels which were very
close in dicots and Gramineae. The latter exhibited, however,
slightly higher values in two cases and a slightly lower value
in one case.
Finally, the case of first codon positions is similar to
that of third codon positions, but values from Gramineae were
not as much higher than dicot values; in one case, that of the
actin gene, the value was even slightly lower.
A conclusion to be drawn from the data of Figs. 2-5 is that
the differences in GC levels previously found in coding
sequences from dicots and Gramineae (1) are also found in the
set of homologous coding sequences under consideration (and in
their codon positions). In other words, the differences were not
due to the gene samples investigated, but were real differences,
5285

Nucleic Acids Research
implying compositional divergences from ancestral genes, caused
by a number of directional mutations. A similar situation was
already described for homologous genes from warm-blooded and
cold-blooded vertebrates(7,8). These results confirm and extend
those previously described by Niesbach-Klosgen et al.(9).
Since the dicots studied in their coding sequences are
characterized by DNAs having relatively low GC levels, whereas
the monocots most investigated belong to some sub-families of
Gramineae that are characterized by high GC values, comparisons
of genes from dicots and monocots, as currently possible, are
strongly biased. If comparisons were done among DNAs (and coding
sequences) from the monocot family Alliaceae and the same
dicots, one would find compositional differences in the opposite

direction.

Codon_usaqe
A detailed statistical analysis of codon usage patterns in
plant genes is being carried out and will be presented elsewhere
in due time. Some conclusions can, however, be already drawn on
the basis of the results obtained in this work.
First of all, as expected, all coding sequences exhibiting
very high GC levels in codon third positions are characterized
by the absence of a very large number (23 to 31) of codons and
by very low levels (RSCU < 0.30) of some additional codons. This
applies to all coding sequences higher than 90t GC in third
codon position. The effect is, however, very evident also in the
80-90% GC range and perhaps already above 70% GC (this cannot be
decided, however, since only one value in the 70-80% GC range is
available). One should expect a similar phenomenon of codon
absence also for very low GC values in third positions. Indeed,
the only GC value in third codon position lower than 30% (18.4%,
in the lignin-forming peroxidase gene from tobacco) shows an
important level of absent or strongly underrepresented codons.
Coding sequences lower than 70% GC in third codon positions
showed a different extent of codon avoidance which was, however,
markedly different for different genes. Indeed, as shown by the
data of Table 1, some genes with 50-60% GC in third codon
position may show a strong codon avoidance (this is the case of
the genes for histone H3, histone H4 and rubisco from dicots),
5286

Nucleic Acids Research
whereas other genes in the same GC range may show no, or almost
no, avoidance of codons (this is the case of chalcone synthase
from dicots and of actin and alcohol dehydrogenase from Z.
may) .
Second, two opposite situations can be found for homologous
genes exhibiting large differences in GC levels of third codon
positions. In the first case, exemplified by the chalcone
synthase genes, the gene exhibiting a very high GC level in
third codon position shows the expected codon avoidance, whereas
its homologous with a low GC level does not. In the second case,
exemplified by rubisco and histones H3 and H4, the genes with
low GC levels exhibit almost the same codon avoidance as their
counterparts with high GC levels. This phenomenon has been seen
for homologous genes from as many as 14 species for the rubisco
gene, suggesting the possible existence of gene-specific
features in codon avoidance. Needless to say, a detailed
statistical analysis is required to allow a more precise
assessment of this situation.
Third, codon avoidance can cover an extremely wide range,
0-52 %, within a single genome, in the case of plants (like Zea
mas and, in all likelihood, of other Gramineae) that have genes
characterized by an extremely broad GC range in third codon
positions. In these genomes an extreme codon avoidance may
concern a very large number of genes. Expectedly codon usage can
also exhibit very large variations within the same genome (Table
1) . This appears to be the case not only for Zea mays, but
generally for plant genomes (as observed for tobacco and
soybean; data not shown). This situation had already been found
for vertebrate genomes (2,10).
Finally, the present results provide additional data
against some misconceptions. Indeed, it is wrong to think that
in compartmentalized genome, like those of plants, "in general,
genes within a taxonomic group exhibit similarities in codon
choice," (11). This situation, first described by Grantham et
al. (12-14) , only applies to compositionally homogeneous
genomes, like those of most bacteria, but not to
compartmentalized genomes (15). For these reasons, pooling codon
usages for dicots and, even more so, for monocots (11) does not
5287

Nucleic Acids Research
make sense. Likewise, concluding that "the relative use of
synonymous codons differs between the monocots and the dicots"
(10) ignores the fact that differences are not due to
differences between codon choices of monocots and dicots, but
mainly to differences between GC-rich genes from Gramineae and
GC-poor genes from the dicots investigated.
CorEarisons of GC content of exons and introns with GC content
of flankig sequences
Previous investigations (1) had indicated an isochore
structure in the nuclear genome of angiosperms. The evidence
rested on the following findings. (i) Extended regions (>
100-200 Kb) around the few genes which were localized on large
DNA fragments fractionated by preparative Cs SO ,/BAMD
centrifugation were shown to be compositionally homogeneous.
Indeed, such fragments were the result of random breakage of DNA
during its preparation, and the gene probed could therefore be
present at any location on the fragments, yet the compositional
distribution of the fragments carrying the gene was within 1l
GC. (ii) The nuclear DNAs of the dicots and of the Gramineae
investigated showed very different compositional patterns. A
major difference was the presence in Gramineae of DNA fragments
that showed a high GC level not represented in the dicot genomes
analyzed. The GC distribution of coding sequences in Gramineae
was similarly highly biased towards high GC values and covered a
GC range that was not overlapped by dicot coding sequences, (a
similar difference was found in introns). Taken together, these
results indicated that GC-rich genes of Gramineae were present
in the GC-rich regions of those genomes, whereas GC-poor genes
corresponded to GC-poor regions.
In the present work, the isochore structure of the nuclear
genome of plants was probed in a different way, namely by
comparing the GC levels of coding sequences and introns of given
genes with the corresponding flanking sequences. A certain level
of variability in the results should be expected for two
reasons: the small size of flanking regions, and the fact that
different relative amounts of 5' and 3' flanking sequences
(which may comprise different amounts of CpG islands; see

5288

Nucleic Acids Research
ref.16) were pooled together. The first problem could be
alleviated, but not eliminated, by using only sequences longer
than 1 Kb. As far as the second problem is concerned, separate
plots of 5' and 3' flanking sequences revealed no significant
deviation from the relationships with pooled flanking sequences.
The correlations found between GC levels of coding
sequences and introns, and GC levels of flanking sequences,
provide yet another evidence for an isochore structure of the
nuclear genomes of plants. In apparent contrast, Brinkmann et
al. (17) claimed that there was "little if any correlation"
between GC levels of introns and flanking sequences, and GC
levels of third codon position. Their data show, however, a
slope of 0.30 and a correlation coefficient of 0.79, which is in
disagreement with their conclusions, but in agreement with ours.
The data on the compositional relationships among GC levels
of exons, introns and flanking sequences lead to some other
conclusions. (i) The results of Fig. 6 suggest that a single
correlation exists between plant genes and flanking sequences
independently of whether dicots or Gramineae are considered.
(ii) The correlations are similar to those found for vertebrate
genomes. The possibility definitely exists, therefore, that a
single general correlation holds in both cases. (iii) As in the
case of vertebrates (2), coding sequences although correlated
with flanking sequences, are systematically higher in GC,
whereas introns are much closer to flanking sequences. (iv) The
different relationships found for first, second and third codon
positions in plants match similar findings apparently valid for
all genomes (15). (v) Flanking sequences of Gramineae exhibit a
higher GC level than those of dicots, in agreement with the
distribution of DNA fragments shown in Fig.l.
ACKNOWLEDGEMENTS
We thank the French Ministry of Foreign Affairs for a
fellowship to G.M., EMBO, Heidelberg (FRG), for a short-term
fellowship to J.S. and INIA for financial support to L.M.M.
*To whom correspondence should be addressed

5289

Nucleic Acids Research
REFERENCES
1. Salinas

2.

3.

4.
5.

6.
7.
8.

9.
10.
11.

and Bernardi G.
J., Matassi G., Montero L.M.
(1988) Nucl. Acids Res. 16:4269-42852.
Bernardi G, Olofsson B., Filipski J., Zerial M., Salinas J.,
(1985)
and Rodier F.
Cuny G, Meunier-Rotival M.
Science 228:953-958.
Gouy M., Milleret F., Mugnier C., Jacobzone M. and Gautier C.
(1984) Nucl. Acids Res. 12:121-127.
Sharp P.M., Tuohy T.M.F. and Mosurcki K.R. (1986)
Nucl. Acids Res. 14:5125-5143.
Ingle J., Pearson G.G. and Sinclair J. (1973) Nature
242:193-197.
Shapiro, H.S. (1976). In: Fasman, G.D. (ed.) Handbook of
Biochemistry and Molecular Biology, 3rd edn, vol. II,
pp. 241-281, CRC Press Inc.
Perrin P. and Bernardi G. (1987) J. Mol. Evol. 26:301-310.
Bernardi G., Mouchiroud D., Gautier C. and Bernardi G. (1988)
J. Mol. Evol. 28: 7-18.
Niesbach-Klosgen, U., Barzen, E., Bernhardt, J., Rohde, W.,
Schwarz-Sommer, Zs., Reif, H.J., Wienand, U. and Saedler, H.
(1987) J. Mol. Evol. 26, 213-225.
Ikemura, T. (1985) Mol. Biol. Evol. 2:13-34.
Murray E.E., Lotzer J. and Eberle M. (1989) Nucl. Acids
Res. 17:477-498.

12. Grantham R., Gautier C. and Gouy M. (1980) Nucl. Acids
Res. 8:1893-1912.
13. Grantham R., Gautier C., Gouy M, Mercier R. and Pave A.
(1980) Nucl. Acids Res. 8:r49-r62.
14. Grantham R., Gautier C., Gouy M, Jacobzone M. and Mercier R.
(1981) Nucl. Acids Res. 9:r43-r74.
15. Bernardi, G. and Bernardi, G. (1986) J. Mol. Evol. 24:1-11.
16. Antequera F. and Bird A. (1988) EMBO J. 8:2295-2299.
17. Brinkmann H., Martinez P., Quigley F., Martin W. and Cerff R.
(1987) J. Mol. Evol. 26:320-328.

5290

Sign up to vote on this title
UsefulNot useful